Federated-trust sharded blockchain for real-time forensics and secure data collaboration in cooperative V2X

Cooperative V2X is evolving toward city-scale deployment, yet current infrastructures still lack a network substrate that jointly provides cross-domain trust, low-latency finality, and privacy-preserving, auditable evidence for safety-critical decisions. This paper proposes a federated-trust sharded blockchain that turns heterogeneous vehicular and roadside measurements into accountable records and enables real-time forensic collaboration and secure data sharing across operators and city management authorities. A federated trust oracle fuses GNSS, OBD, IMU, RSU observations, and device attestations into uncertainty-aware scores that steer committee election, voting weights, and traffic shaping in each shard. On this basis, we design a hybrid cross-shard commit protocol with adaptive finality, combining atomic channels for forensic-critical transactions and optimistic channels for routine collaboration, and we establish safety/liveness conditions and provide proof sketches under the stated partial-synchrony assumptions and bounded collusion. For the forensic layer, a two-stage pipeline anchors minimal sufficient evidence with sub-second local finality, while editable proofs built on traffic-aware extended Merkle trees and zero-knowledge attestations support publicly verifiable, legally compliant edits with O(log n) verification overhead. An SLA-aware, learning-assisted scheduler adapts committee size, batching, and cross-shard parallelism to dynamic traffic and attack patterns so as to meet latency, throughput, and rollback targets. Large-topology containerized emulation on a dedicated workstation, complemented by a small hardware-in-the-loop testbed, shows that the proposed framework achieves sub-second forensic anchoring and 95th-percentile cross-shard finality below $$1.2$$ s. Across the representative baselines used in this study, it improves effective throughput by up to $$35%$$; in particular, at comparable Lₚ₉₅, it achieves $$1.6$$–2.3× higher TPS than the single-chain HotStuff baseline under the tested emulation conditions, while reducing rollback rate and per-event bandwidth by up to $$40%$$ and $$25%$$, respectively. These results indicate that the proposed system can shorten incident response, strengthen accountability in crash investigations and recalls, and provide a practical foundation for privacy-preserving, transparent data collaboration between mobility operators and urban management departments.